We propose a transfer-learning multi-input multi-output (TL-MIMO) scheme to significantly reduce the required training complexity for converging the equalizers in mode-division multiplexing (MDM) systems. Based on a built three-mode (LP01, LP11a and LP11b) multiplexed experimental system, we thoughtfully investigate the TL-MIMO performances on the three-typed data, collecting from different sampling times, launched optical powers, and input optical signal-to-noise ratios (OSNRs). The dramatic reduction of 40%~83.33% on the required training complexity is achieved in all of three scenarios. Furthermore, the good stability of TL-MIMO in both the launched power and OSNR test bands has also been proved.